Why smart statistics are the key to fighting crime
-
0:01 - 0:03In 2007, I became the attorney general
-
0:03 - 0:05of the state of New Jersey.
-
0:05 - 0:07Before that, I'd been a criminal prosecutor,
-
0:07 - 0:10first in the Manhattan district attorney's office,
-
0:10 - 0:13and then at the United States Department of Justice.
-
0:13 - 0:15But when I became the attorney general,
-
0:15 - 0:19two things happened that changed
the way I see criminal justice. -
0:19 - 0:21The first is that I asked what I thought
-
0:21 - 0:23were really basic questions.
-
0:23 - 0:26I wanted to understand who we were arresting,
-
0:26 - 0:28who we were charging,
-
0:28 - 0:30and who we were putting in our nation's jails
-
0:30 - 0:31and prisons.
-
0:31 - 0:33I also wanted to understand
-
0:33 - 0:34if we were making decisions
-
0:34 - 0:37in a way that made us safer.
-
0:37 - 0:40And I couldn't get this information out.
-
0:40 - 0:43It turned out that most big criminal justice agencies
-
0:43 - 0:45like my own
-
0:45 - 0:47didn't track the things that matter.
-
0:47 - 0:50So after about a month of being incredibly frustrated,
-
0:50 - 0:52I walked down into a conference room
-
0:52 - 0:54that was filled with detectives
-
0:54 - 0:57and stacks and stacks of case files,
-
0:57 - 0:58and the detectives were sitting there
-
0:58 - 1:00with yellow legal pads taking notes.
-
1:00 - 1:02They were trying to get the information
-
1:02 - 1:03I was looking for
-
1:03 - 1:05by going through case by case
-
1:05 - 1:07for the past five years.
-
1:07 - 1:09And as you can imagine,
-
1:09 - 1:11when we finally got the results, they weren't good.
-
1:11 - 1:13It turned out that we were doing
-
1:13 - 1:15a lot of low-level drug cases
-
1:15 - 1:16on the streets just around the corner
-
1:16 - 1:19from our office in Trenton.
-
1:19 - 1:20The second thing that happened
-
1:20 - 1:24is that I spent the day in the Camden,
New Jersey police department. -
1:24 - 1:26Now, at that time, Camden, New Jersey,
-
1:26 - 1:28was the most dangerous city in America.
-
1:28 - 1:32I ran the Camden Police
Department because of that. -
1:32 - 1:34I spent the day in the police department,
-
1:34 - 1:37and I was taken into a room
with senior police officials, -
1:37 - 1:39all of whom were working hard
-
1:39 - 1:42and trying very hard to reduce crime in Camden.
-
1:42 - 1:44And what I saw in that room,
-
1:44 - 1:46as we talked about how to reduce crime,
-
1:46 - 1:50were a series of officers with a
lot of little yellow sticky notes. -
1:50 - 1:53And they would take a yellow sticky
and they would write something on it -
1:53 - 1:55and they would put it up on a board.
-
1:55 - 1:57And one of them said,
"We had a robbery two weeks ago. -
1:57 - 1:59We have no suspects."
-
1:59 - 2:04And another said, "We had a shooting in this neighborhood last week. We have no suspects."
-
2:04 - 2:06We weren't using data-driven policing.
-
2:06 - 2:08We were essentially trying to fight crime
-
2:08 - 2:11with yellow Post-it notes.
-
2:11 - 2:13Now, both of these things made me realize
-
2:13 - 2:16fundamentally that we were failing.
-
2:16 - 2:19We didn't even know who was
in our criminal justice system, -
2:19 - 2:22we didn't have any data about
the things that mattered, -
2:22 - 2:25and we didn't share data or use analytics
-
2:25 - 2:27or tools to help us make better decisions
-
2:27 - 2:29and to reduce crime.
-
2:29 - 2:31And for the first time, I started to think
-
2:31 - 2:33about how we made decisions.
-
2:33 - 2:35When I was an assistant D.A.,
-
2:35 - 2:37and when I was a federal prosecutor,
-
2:37 - 2:38I looked at the cases in front of me,
-
2:38 - 2:41and I generally made decisions based on my instinct
-
2:41 - 2:43and my experience.
-
2:43 - 2:44When I became attorney general,
-
2:44 - 2:46I could look at the system as a whole,
-
2:46 - 2:48and what surprised me is that I found
-
2:48 - 2:50that that was exactly how we were doing it
-
2:50 - 2:52across the entire system --
-
2:52 - 2:54in police departments, in prosecutors's offices,
-
2:54 - 2:57in courts and in jails.
-
2:57 - 2:59And what I learned very quickly
-
2:59 - 3:03is that we weren't doing a good job.
-
3:03 - 3:05So I wanted to do things differently.
-
3:05 - 3:07I wanted to introduce data and analytics
-
3:07 - 3:09and rigorous statistical analysis
-
3:09 - 3:11into our work.
-
3:11 - 3:14In short, I wanted to moneyball criminal justice.
-
3:14 - 3:16Now, moneyball, as many of you know,
-
3:16 - 3:17is what the Oakland A's did,
-
3:17 - 3:19where they used smart data and statistics
-
3:19 - 3:21to figure out how to pick players
-
3:21 - 3:22that would help them win games,
-
3:22 - 3:25and they went from a system that
was based on baseball scouts -
3:25 - 3:27who used to go out and watch players
-
3:27 - 3:29and use their instinct and experience,
-
3:29 - 3:31the scouts' instincts and experience,
-
3:31 - 3:32to pick players, from one to use
-
3:32 - 3:35smart data and rigorous statistical analysis
-
3:35 - 3:38to figure out how to pick players
that would help them win games. -
3:38 - 3:40It worked for the Oakland A's,
-
3:40 - 3:42and it worked in the state of New Jersey.
-
3:42 - 3:45We took Camden off the top of the list
-
3:45 - 3:47as the most dangerous city in America.
-
3:47 - 3:50We reduced murders there by 41 percent,
-
3:50 - 3:53which actually means 37 lives were saved.
-
3:53 - 3:57And we reduced all crime in the city by 26 percent.
-
3:57 - 4:00We also changed the way
we did criminal prosecutions. -
4:00 - 4:02So we went from doing low-level drug crimes
-
4:02 - 4:03that were outside our building
-
4:03 - 4:06to doing cases of statewide importance,
-
4:06 - 4:09on things like reducing violence
with the most violent offenders, -
4:09 - 4:11prosecuting street gangs,
-
4:11 - 4:14gun and drug trafficking, and political corruption.
-
4:14 - 4:17And all of this matters greatly,
-
4:17 - 4:19because public safety to me
-
4:19 - 4:21is the most important function of government.
-
4:21 - 4:24If we're not safe, we can't be educated,
-
4:24 - 4:25we can't be healthy,
-
4:25 - 4:28we can't do any of the other things
we want to do in our lives. -
4:28 - 4:30And we live in a country today
-
4:30 - 4:33where we face serious criminal justice problems.
-
4:33 - 4:36We have 12 million arrests every single year.
-
4:36 - 4:38The vast majority of those arrests
-
4:38 - 4:41are for low-level crimes, like misdemeanors,
-
4:41 - 4:4370 to 80 percent.
-
4:43 - 4:45Less than five percent of all arrests
-
4:45 - 4:47are for violent crime.
-
4:47 - 4:49Yet we spend 75 billion,
-
4:49 - 4:50that's b for billion,
-
4:50 - 4:55dollars a year on state and local corrections costs.
-
4:55 - 4:57Right now, today, we have 2.3 million people
-
4:57 - 4:59in our jails and prisons.
-
4:59 - 5:02And we face unbelievable public safety challenges
-
5:02 - 5:04because we have a situation
-
5:04 - 5:07in which two thirds of the people in our jails
-
5:07 - 5:09are there waiting for trial.
-
5:09 - 5:11They haven't yet been convicted of a crime.
-
5:11 - 5:13They're just waiting for their day in court.
-
5:13 - 5:17And 67 percent of people come back.
-
5:17 - 5:20Our recidivism rate is amongst
the highest in the world. -
5:20 - 5:22Almost seven in 10 people who are released
-
5:22 - 5:23from prison will be rearrested
-
5:23 - 5:27in a constant cycle of crime and incarceration.
-
5:27 - 5:30So when I started my job at the Arnold Foundation,
-
5:30 - 5:33I came back to looking at a lot of these questions,
-
5:33 - 5:34and I came back to thinking about how
-
5:34 - 5:37we had used data and analytics to transform
-
5:37 - 5:39the way we did criminal justice in New Jersey.
-
5:39 - 5:41And when I look at the criminal justice system
-
5:41 - 5:43in the United States today,
-
5:43 - 5:45I feel the exact same way that I did
-
5:45 - 5:47about the state of New Jersey when I started there,
-
5:47 - 5:50which is that we absolutely have to do better,
-
5:50 - 5:52and I know that we can do better.
-
5:52 - 5:54So I decided to focus
-
5:54 - 5:56on using data and analytics
-
5:56 - 5:59to help make the most critical decision
-
5:59 - 6:00in public safety,
-
6:00 - 6:02and that decision is the determination
-
6:02 - 6:05of whether, when someone has been arrested,
-
6:05 - 6:07whether they pose a risk to public safety
-
6:07 - 6:08and should be detained,
-
6:08 - 6:10or whether they don't pose a risk to public safety
-
6:10 - 6:12and should be released.
-
6:12 - 6:14Everything that happens in criminal cases
-
6:14 - 6:16comes out of this one decision.
-
6:16 - 6:17It impacts everything.
-
6:17 - 6:19It impacts sentencing.
-
6:19 - 6:21It impacts whether someone gets drug treatment.
-
6:21 - 6:23It impacts crime and violence.
-
6:23 - 6:25And when I talk to judges around the United States,
-
6:25 - 6:27which I do all the time now,
-
6:27 - 6:29they all say the same thing,
-
6:29 - 6:32which is that we put dangerous people in jail,
-
6:32 - 6:35and we let non-dangerous, nonviolent people out.
-
6:35 - 6:37They mean it and they believe it.
-
6:37 - 6:39But when you start to look at the data,
-
6:39 - 6:42which, by the way, the judges don't have,
-
6:42 - 6:43when we start to look at the data,
-
6:43 - 6:46what we find time and time again,
-
6:46 - 6:48is that this isn't the case.
-
6:48 - 6:49We find low-risk offenders,
-
6:49 - 6:53which makes up 50 percent of our
entire criminal justice population, -
6:53 - 6:55we find that they're in jail.
-
6:55 - 6:58Take Leslie Chew, who was a Texas man
-
6:58 - 7:01who stole four blankets on a cold winter night.
-
7:01 - 7:03He was arrested, and he was kept in jail
-
7:03 - 7:05on 3,500 dollars bail,
-
7:05 - 7:08an amount that he could not afford to pay.
-
7:08 - 7:11And he stayed in jail for eight months
-
7:11 - 7:13until his case came up for trial,
-
7:13 - 7:17at a cost to taxpayers of more than 9,000 dollars.
-
7:17 - 7:19And at the other end of the spectrum,
-
7:19 - 7:21we're doing an equally terrible job.
-
7:21 - 7:23The people who we find
-
7:23 - 7:25are the highest-risk offenders,
-
7:25 - 7:27the people who we think have the highest likelihood
-
7:27 - 7:29of committing a new crime if they're released,
-
7:29 - 7:32we see nationally that 50 percent of those people
-
7:32 - 7:34are being released.
-
7:34 - 7:37The reason for this is the way we make decisions.
-
7:37 - 7:39Judges have the best intentions
-
7:39 - 7:41when they make these decisions about risk,
-
7:41 - 7:43but they're making them subjectively.
-
7:43 - 7:46They're like the baseball scouts 20 years ago
-
7:46 - 7:48who were using their instinct and their experience
-
7:48 - 7:50to try to decide what risk someone poses.
-
7:50 - 7:52They're being subjective,
-
7:52 - 7:55and we know what happens
with subjective decision making, -
7:55 - 7:58which is that we are often wrong.
-
7:58 - 7:59What we need in this space
-
7:59 - 8:02are strong data and analytics.
-
8:02 - 8:03What I decided to look for
-
8:03 - 8:06was a strong data and analytic risk assessment tool,
-
8:06 - 8:09something that would let judges actually understand
-
8:09 - 8:11with a scientific and objective way
-
8:11 - 8:13what the risk was that was posed
-
8:13 - 8:14by someone in front of them.
-
8:14 - 8:16I looked all over the country,
-
8:16 - 8:18and I found that between five and 10 percent
-
8:18 - 8:19of all U.S. jurisdictions
-
8:19 - 8:22actually use any type of risk assessment tool,
-
8:22 - 8:24and when I looked at these tools,
-
8:24 - 8:26I quickly realized why.
-
8:26 - 8:29They were unbelievably expensive to administer,
-
8:29 - 8:30they were time-consuming,
-
8:30 - 8:32they were limited to the local jurisdiction
-
8:32 - 8:34in which they'd been created.
-
8:34 - 8:35So basically, they couldn't be scaled
-
8:35 - 8:38or transferred to other places.
-
8:38 - 8:40So I went out and built a phenomenal team
-
8:40 - 8:42of data scientists and researchers
-
8:42 - 8:43and statisticians
-
8:43 - 8:46to build a universal risk assessment tool,
-
8:46 - 8:49so that every single judge in
the United States of America -
8:49 - 8:53can have an objective, scientific measure of risk.
-
8:53 - 8:55In the tool that we've built,
-
8:55 - 8:58what we did was we collected 1.5 million cases
-
8:58 - 8:59from all around the United States,
-
8:59 - 9:01from cities, from counties,
-
9:01 - 9:02from every single state in the country,
-
9:02 - 9:04the federal districts.
-
9:04 - 9:06And with those 1.5 million cases,
-
9:06 - 9:08which is the largest data set on pretrial
-
9:08 - 9:10in the United States today,
-
9:10 - 9:12we were able to basically find that there were
-
9:12 - 9:15900-plus risk factors that we could look at
-
9:15 - 9:18to try to figure out what mattered most.
-
9:18 - 9:20And we found that there were nine specific things
-
9:20 - 9:22that mattered all across the country
-
9:22 - 9:25and that were the most highly predictive of risk.
-
9:25 - 9:29And so we built a universal risk assessment tool.
-
9:29 - 9:31And it looks like this.
-
9:31 - 9:33As you'll see, we put some information in,
-
9:33 - 9:35but most of it is incredibly simple,
-
9:35 - 9:37it's easy to use,
-
9:37 - 9:40it focuses on things like the
defendant's prior convictions, -
9:40 - 9:42whether they've been sentenced to incarceration,
-
9:42 - 9:44whether they've engaged in violence before,
-
9:44 - 9:46whether they've even failed to come back to court.
-
9:46 - 9:49And with this tool, we can predict three things.
-
9:49 - 9:51First, whether or not someone will commit
-
9:51 - 9:52a new crime if they're released.
-
9:52 - 9:54Second, for the first time,
-
9:54 - 9:56and I think this is incredibly important,
-
9:56 - 9:58we can predict whether someone will commit
-
9:58 - 9:59an act of violence if they're released.
-
9:59 - 10:01And that's the single most important thing
-
10:01 - 10:03that judges say when you talk to them.
-
10:03 - 10:05And third, we can predict whether someone
-
10:05 - 10:07will come back to court.
-
10:07 - 10:10And every single judge in the
United States of America can use it, -
10:10 - 10:14because it's been created on a universal data set.
-
10:14 - 10:16What judges see if they run the risk assessment tool
-
10:16 - 10:19is this -- it's a dashboard.
-
10:19 - 10:21At the top, you see the New Criminal Activity Score,
-
10:21 - 10:23six of course being the highest,
-
10:23 - 10:26and then in the middle you
see, "Elevated risk of violence." -
10:26 - 10:27What that says is that this person
-
10:27 - 10:30is someone who has an elevated risk of violence
-
10:30 - 10:31that the judge should look twice at.
-
10:31 - 10:33And then, towards the bottom,
-
10:33 - 10:35you see the Failure to Appear Score,
-
10:35 - 10:36which again is the likelihood
-
10:36 - 10:39that someone will come back to court.
-
10:39 - 10:41Now I want to say something really important.
-
10:41 - 10:44It's not that I think we should be eliminating
-
10:44 - 10:46the judge's instinct and experience
-
10:46 - 10:48from this process.
-
10:48 - 10:49I don't.
-
10:49 - 10:51I actually believe the problem that we see
-
10:51 - 10:54and the reason that we have
these incredible system errors, -
10:54 - 10:57where we're incarcerating
low-level, nonviolent people -
10:57 - 11:00and we're releasing high-risk, dangerous people,
-
11:00 - 11:03is that we don't have an objective measure of risk.
-
11:03 - 11:04But what I believe should happen
-
11:04 - 11:07is that we should take that
data-driven risk assessment -
11:07 - 11:10and combine that with the
judge's instinct and experience -
11:10 - 11:13to lead us to better decision making.
-
11:13 - 11:16The tool went statewide in Kentucky on July 1,
-
11:16 - 11:20and we're about to go up in a
number of other U.S. jurisdictions. -
11:20 - 11:22Our goal, quite simply, is that every single judge
-
11:22 - 11:24in the United States will use a data-driven risk tool
-
11:24 - 11:26within the next five years.
-
11:26 - 11:28We're now working on risk tools
-
11:28 - 11:31for prosecutors and for police officers as well,
-
11:31 - 11:34to try to take a system that runs today
-
11:34 - 11:37in America the same way it did 50 years ago,
-
11:37 - 11:39based on instinct and experience,
-
11:39 - 11:41and make it into one that runs
-
11:41 - 11:43on data and analytics.
-
11:43 - 11:45Now, the great news about all this,
-
11:45 - 11:47and we have a ton of work left to do,
-
11:47 - 11:48and we have a lot of culture to change,
-
11:48 - 11:50but the great news about all of it
-
11:50 - 11:52is that we know it works.
-
11:52 - 11:54It's why Google is Google,
-
11:54 - 11:57and it's why all these baseball teams use moneyball
-
11:57 - 11:58to win games.
-
11:58 - 12:00The great news for us as well
-
12:00 - 12:02is that it's the way that we can transform
-
12:02 - 12:04the American criminal justice system.
-
12:04 - 12:07It's how we can make our streets safer,
-
12:07 - 12:09we can reduce our prison costs,
-
12:09 - 12:11and we can make our system much fairer
-
12:11 - 12:13and more just.
-
12:13 - 12:15Some people call it data science.
-
12:15 - 12:17I call it moneyballing criminal justice.
-
12:17 - 12:19Thank you.
-
12:19 - 12:23(Applause)
- Title:
- Why smart statistics are the key to fighting crime
- Speaker:
- Anne Milgram
- Description:
-
When she became the attorney general of New Jersey in 2007, Anne Milgram quickly discovered a few startling facts: not only did her team not really know who they were putting in jail, but they had no way of understanding if their decisions were actually making the public safer. And so began her ongoing, inspirational quest to bring data analytics and statistical analysis to the US criminal justice system.
- Video Language:
- English
- Team:
- closed TED
- Project:
- TEDTalks
- Duration:
- 12:41
janet dragojevic edited English subtitles for Why smart statistics are the key to fighting crime | ||
Morton Bast approved English subtitles for Why smart statistics are the key to fighting crime | ||
Morton Bast edited English subtitles for Why smart statistics are the key to fighting crime | ||
Morton Bast edited English subtitles for Why smart statistics are the key to fighting crime | ||
Madeleine Aronson accepted English subtitles for Why smart statistics are the key to fighting crime | ||
Madeleine Aronson edited English subtitles for Why smart statistics are the key to fighting crime | ||
Madeleine Aronson edited English subtitles for Why smart statistics are the key to fighting crime | ||
Joseph Geni edited English subtitles for Why smart statistics are the key to fighting crime |